import pickle

import matplotlib.pyplot as plt
import matplotlib as mpl

import numpy as np
%matplotlib inline
import pandas as pd

import os, glob
import seaborn as sns
sns.set(style="darkgrid")

from IPython.display import display, HTML, Image

import plotly
import plotly.express as px
import plotly.graph_objects as go

data5H =pd.read_csv('15_PCc_02_LHS_500_54854_01_s1_G.csv')
data5K =pd.read_csv('15_PCc_02_LHS_5000_54854_01_s1_G.csv')
#data50K =pd.read_csv('15_PCc_02_LHS_50000_54854_01_s1_G.csv')
#data5k =pd.read_csv(r'F:\ML4ChemcialReactions\chemical5K\10_PC_02_LHS_5000_54854_01_s1_G.csv')

### Remove the space from the column names
data5H.columns =[col.strip() for col in data5H.columns]
data5K.columns =[col.strip() for col in data5K.columns]
#data50K.columns =[col.strip() for col in data50K.columns]

data5H.columns

Index(['T', 'CaO', 'SiO2', 'CO2', 'H2O', 'pH', 'MassWater', 'Ca_aq', 'Si_aq',
       'C_aq', 'O_aq', 'H_aq', 'Ca_s', 'Si_s', 'C_s', 'O_s', 'H_s',
       'Portlandite', 'AmorfSi', 'Calcite', 'mCSHQ', 'Ca_ss', 'Si_ss',
       'H2O_ss', 'V_s', 'Gel_water'],
      dtype='object')

#data50K.describe().T

pureMinerals = ['Portlandite', 'AmorfSi', 'Gibbsite', 'Katoite', 'Monosulfate', 'Ettringite', 'Straetlingite', 'Chabazite', 'Calcite', 'Hemicarbonate', 'Monocarbonate', 'Thaumasite'] # 'Gypsum',

data5K.columns

Index(['T', 'CaO', 'SiO2', 'CO2', 'H2O', 'pH', 'MassWater', 'Ca_aq', 'Si_aq',
       'C_aq', 'O_aq', 'H_aq', 'Ca_s', 'Si_s', 'C_s', 'O_s', 'H_s',
       'Portlandite', 'AmorfSi', 'Calcite', 'mCSHQ', 'Ca_ss', 'Si_ss',
       'H2O_ss', 'V_s', 'Gel_water'],
      dtype='object')

Label the dataset based on the minerals

for mineral in ['Portlandite', 'AmorfSi', 'Calcite']:
    conditions = [
         (data5K[mineral]>0),
         (data5K[mineral]==0)] 
    values = [mineral, 'No'+mineral]
    data5K[mineral+'_stat'] = np.select(conditions, values)    
         
    

colStats= [col for col in data5K.columns if '_stat' in col]
data5K['group'] = data5K[colStats].agg('-'.join, axis=1)

data5K['group'].unique()

array(['Portlandite-NoAmorfSi-Calcite', 'NoPortlandite-AmorfSi-Calcite',
       'NoPortlandite-NoAmorfSi-Calcite'], dtype=object)

Supervised grouping with ML models

from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import train_test_split

# metrics are used to find accuracy or error
from sklearn import metrics 

data5K.columns

Index(['T', 'CaO', 'SiO2', 'CO2', 'H2O', 'pH', 'MassWater', 'Ca_aq', 'Si_aq',
       'C_aq', 'O_aq', 'H_aq', 'Ca_s', 'Si_s', 'C_s', 'O_s', 'H_s',
       'Portlandite', 'AmorfSi', 'Calcite', 'mCSHQ', 'Ca_ss', 'Si_ss',
       'H2O_ss', 'V_s', 'Gel_water', 'Portlandite_stat', 'AmorfSi_stat',
       'Calcite_stat', 'group'],
      dtype='object')

inputVars= ['CaO', 'SiO2', 'CO2', 'H2O']
targVars = ['group']

X = data5K[inputVars]
y = data5K[targVars]

# i.e. 80 % training dataset and 20 % test datasets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.20)

# creating a RF classifier
clf = RandomForestClassifier(n_estimators = 500) 
 
# Training the model on the training dataset
# fit function is used to train the model using the training sets as parameters
clf.fit(X_train, y_train)
 
# performing predictions on the test dataset
y_pred = clf.predict(X_test)
 

print("ACCURACY OF THE MODEL: ", metrics.accuracy_score(y_test, y_pred))

C:\Users\Cyang.MDM-MOD1-D1\AppData\Local\Temp\ipykernel_3932\337620559.py:6: DataConversionWarning: A column-vector y was passed when a 1d array was expected. Please change the shape of y to (n_samples,), for example using ravel().
  clf.fit(X_train, y_train)

ACCURACY OF THE MODEL:  0.963

X_test =pd.concat([X_test,y_test],axis=1)
X_test['pred']=y_pred

X_test.columns

Index(['CaO', 'SiO2', 'CO2', 'H2O', 'group', 'pred'], dtype='object')

fig =px.scatter_ternary(X_test, a="CaO", b="SiO2", c="CO2",color='group')
fig.show()

fig =px.scatter_ternary(X_test, a="CaO", b="SiO2", c="CO2",color='pred')
fig.show()